Open fmcclean opened 4 years ago
Thinking about this further, is it ever necessary to run workflows more than once? If not, is the concept of workflow instances required in the UI or would it be simpler to link the status to the workflow itself?
Some pretty fundamental questions here! Really interested in getting some more info. Could you explain a bit more the way that you're using the workflows at the moment and why you never expect to run a workflow more than once?
Since model parameters are fixed within each workflow, the output data will always be the same and therefore it doesn't make sense to run the same workflow twice. If there is a random component to a model then it needs to be seeded with a parameter otherwise the results aren't reproducible.
Hi Fergus - it is actually possible to change your parameters, then execute that workflow without saving it. This creates a new instance, but no new workflow - the record of the changed parameter is then created in the instance.
This does seem to lend more weight to your suggestion of supporting named instances, so that you can have a quick reference to the parameters in the title.
Captured in ticket CORE-1407 (DAFNI Internal)
Thanks for the update @BethanPerkins - I wasn't aware that you could do this. How do I go about it?
I've just noticed the execute button which I guess is how you do it. It could do with being a primary button in my opinion as this option isn't very obvious. It might be helpful to call it 'Execute without saving'.
If workflows are run more than once (presumably this would only be the case for Monte Carlo analysis) then the instances are indistinguishable